Spectrogram analysis of electrocardiogram with normal sinus rhythm, arrhythmia and atrial fibrillation

This paper intends to identify a method to distinguish between ECG signals from the Normal Sinus Rhythm (NSR) database and those from the Arrhythmia database. It also attempts to distinguish between NSR and Atrial Fibrillation (Afib). The ECG samples were taken from the database of MIT-BIH. For the first experiment, three-second segments of the entire ECG signals of NSR and that of the P wave alone were analyzed using their power spectral density (PSD). These were then compared to the ECG and P wave of those with Arrhythmia. Spectrum of the P wave clearly differentiates NSR and Arrhythmia. NSR signals exhibit a band within the 10 Hz region while ECG with arrhythmia has a frequency band less than 10 Hz. Furthermore, the dominant frequencies of the NSR extend up to 10 Hz while that of an arrhythmic ECG has very low dominant frequencies. For the second experiment, NSR and Afib signals were cut into 10-second segments. The Afib signals contain annotation marking episodes of normal rhythm and atrial fibrillation. NSR signals show significant frequency content up to more than 30 Hz while majority of the normal rhythm episodes from Afib had significant spectral content concentrated around 4 Hz with lower spectral density than that of NSR.